Non-Gaussian background modeling for anomaly detection in hyperspectral images

Abstract

In this paper, we address the problem of unsupervised detection of anomalies in hyperspectral images. Our proposed method is based on a novel statistical background modeling approach that combines local and global approaches and does not assume Gaussianity. The local-global background model has the ability to adapt to all nuances of the background process, like local models, but avoids overfitting that may result due a too high number of degrees of freedom, producing a high false alarm rate. This is achieved by globally combining the local background models into a “dictionary”, which serves to remove false alarms. Experimental results strongly prove the effectiveness of the proposed algorithm. These results show that the proposed local-global algorithm performs better than several other local or global anomaly detection techniques, such as the well known RX or its Gaussian Mixture version (GMM-RX).